How a Loop Audit Works (Step by Step)
Many people think better results come from changing settings faster.
In reality, most problems come from misinterpreting the data.
Before changing anything, I look for patterns.
Step 1: Collect the data
I usually review:
- 3–7 days of CGM data
- AAPS/Trio settings
- Meal entries
- Bolus history
- SMB activity
- Nighttime trends
- Notes about exercise, stress, illness, or unusual events
The goal is to separate real patterns from random diabetes noise.
Step 2: Identify the actual problem
A high after lunch doesn't automatically mean the carb ratio is wrong.
A nighttime low doesn't automatically mean basal is too high.
Many issues are caused by:
- Prebolus timing
- Meal composition
- Insulin action profile
- SMB behavior
- Site absorption issues
- Counterregulation effects
Finding the true cause is often more important than changing settings.
Step 3: Build a hypothesis
Instead of guessing, I create a working theory based on the data.
Examples:
- "The prebolus is too short."
- "SMBs are arriving too aggressively."
- "The insulin model doesn't match the actual insulin."
- "Counterregulation is being mistaken for a settings problem."
Step 4: Test one variable at a time
The fastest way to get lost is changing five things at once.
I prefer controlled testing with clean data whenever possible.
One change.
One observation.
One conclusion.
Step 5: Review the results
After testing, we compare:
- Before
- After
- What improved
- What did not improve
Then we decide on the next step.
What makes this different?
I don't promise perfect graphs.
Diabetes will always find ways to surprise us.
What I do promise is a structured diagnostic process focused on finding causes rather than chasing symptoms.
Over the years I've learned that the biggest improvements often come from identifying things that were previously overlooked or incorrectly interpreted.
If you're struggling with unexplained highs, lows, overnight instability, SMB behavior, meal responses, or recurring patterns that don't make sense, this process may help uncover what's really happening behind the data.